On Sat, 10 Oct 2020 16:52:48 -0700 John Fastabend wrote: > Jakub Kicinski wrote: > > FWIW I took a quick swing at testing it with the HW I have and it did > > exactly what hardware should do. The TX unit entered an error state > > and then the driver detected that and reset it a few seconds later. > > Ths seems like the right thing to do in my opinion. If the > stack gives the NIC garbage entering error state and reset > sounds expected. Thanks for actually trying it by the way. > > We might have come to different conclusions though from my side > the conclusion is, good nothing horrible happened no MTU check needed. > If the user spews garbage at the nic from the BPF program great it > gets dropped and causes the driver/nic to punish you a bit by staying > hung. Fix your BPF program.
Right probably difference of perspective. I understand that from Cilium's POV you can probably feel pretty confident about the BPF programs that are running. I bet Maciej is even more confident with Android! But in principle BPF was supposed to make the kernel end user programmable. We have to ensure it's safe. > > > > And all this for what? Saving 2 cycles on a branch that will almost > > > > never be taken? > > 2 cycles here and 2 cycles there .... plus complexity to think about > it. Eventually it all adds up. At the risk of entering bike shedding > territory maybe. Not sure it's a bike shedding territory but I doubt you want to be making either the complexity or the performance argument to a fellow TLS maintainer.. cough cough.. ;)
